in re credit suisse_304_defendant's brief re motion to preclude expert opinions

7/27/2019 In Re Credit Suisse_304_Defendant's Brief Re Motion to Preclude Expert Opinions

1/61

UNITED STATES DISTRICT COURTDISTRICT OF MASSACHUSETTS

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - x

IN RE CREDIT SUISSE AOL

SECURITIES LITIGATION

This document relates to:

ALL ACTIONS

))))))))))

Case No. 1:02 CV 12146(Judge Gertner)

Oral Argument Requested

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - x

MEMORANDUM IN SUPPORT OF THE MOTION OF

DEFENDANTS CREDIT SUISSE SECURITIES (USA) LLC,

CREDIT SUISSE (USA), INC., FRANK QUATTRONE,

ELLIOTT ROGERS, JAMIE KIGGEN AND LAURA MARTIN

TO PRECLUDE THE EXPERT OPINIONS OF

SCOTT D. HAKALA, M. LAURENTIUS MARAIS, BERNARD

BLACK AND REINIER KRAAKMAN

Lawrence Portnoy (admittedpro hac vice)Avi Gesser (admittedpro hac vice)DAVIS POLK & WARDWELL450 Lexington AvenueNew York, New York 10017Tel: (212) 450-4000

Fax: (212) 450-3800

Attorneys for Defendants Credit Suisse

Securities (USA) LLC and Credit Suisse (USA),

Inc.

Robert A. Buhlman (BBO #554393)Siobhan E. Mee (BBO #640372)BINGHAM MCCUTCHEN LLPOne Federal StreetBoston, Massachusetts 02110Tel: (617) 951-8000

Fax: (617) 951-8736

Attorneys for Defendants Credit Suisse

Securities (USA) LLC, Credit Suisse (USA),

Inc., Elliott Rogers and Jamie Kiggen

Case 1:02-cv-12146-NMG Document 304 Filed 04/21/09 Page 1 of 61


2/61

James R. Carroll (BBO #554426)Eben P. Colby (BBO #651456)SKADDEN, ARPS, SLATE, MEAGHER &FLOM LLPOne Beacon Street

Boston, Massachusetts 02108Tel: (617) 573-4800Fax: (617) 573-4822

Attorneys for Defendant Laura Martin

Jeff E. Butler (admittedpro hac vice)CLIFFORD CHANCE LLP31 West 52nd StreetNew York, New York 10019

Tel: (212) 878-8000Fax: (212) 878-8375

Warren L. Feldman (admittedpro hac vice)SKADDEN, ARPS, SLATE, MEAGHER &FLOM LLPFour Times SquareNew York, New York 10036Tel: (212) 735-2420Fax: (917) 777-2420

Attorneys for Defendant Elliott Rogers

Henry Putzel, III (admittedpro hac vice)LAW OFFICE OF HENRY PUTZEL III565 Fifth AvenueNew York, New York 10017Tel: (212) 661-0066

Fax: (212) 661-0415

Attorney for Defendant Jamie Kiggen

Jeffrey B. Rudman (BBO #433380)Jonathan A. Shapiro (BBO #567838)Timothy J. Perla (BBO #660447)WILMER CUTLER PICKERING HALEAND DORR LLP60 State StreetBoston, Massachusetts 02109

Tel: (617) 526-6000Fax: (617) 526-5000

Kenneth G. Hausman (admittedpro hac vice)Barbara A. Winters (admittedpro hac vice)HOWARD, RICE, NEMEROVSKI,CANADY, FALK & RABKINA Professional CorporationThree Embarcadero Center, Seventh FloorSan Francisco, California 94111-4024Tel: (415) 434-1600

Fax: (415) 217-5910

Attorneys for Defendant Frank Quattrone



3/61

TABLE OF CONTENTS

PAGE

PRELIMINARY STATEMENT .................................................................................................... 1

ARGUMENT.................................................................................................................................. 4

I. Dr. Hakalas Opinions Are Entirely Unreliable and Must Be Precluded ........................... 7

A. Dr. Hakalas Event Study Fails to Control for Confounding Events andContravenes Market Efficiency .............................................................................. 8

1. Dr. Hakalas Failure to Control for Confounding InformationRenders His Event Study Unreliable .......................................................... 8

2. Dr. Hakalas Disregard of Prior Disclosures Contradicts MarketEfficiency and Renders His Event Study Unreliable................................ 11

B. Dr. Hakalas Event Study Methodology Is Fundamentally Flawed ..................... 12

1. Dr. Hakala Has Failed to Use Relevant Event Dates and HasManufactured the Relevance of the Events He SupposedlyTested........................................................................................................ 14

2. Dr. Hakala Continues to Improperly Use Dummy Variables,Which Renders His Methodology Impossible to Replicate...................... 20

(a) Dr. Hakalas Continued Overuse of Dummy VariablesRequires His Exclusion................................................................. 20

(b) Dr. Hakalas Approach to Dummy Variables Cannot BeReplicated by Other Economists or Even Dr. Hakala................... 24

C. Dr. Hakalas Use of a Proxy for the Effect of Defendants Statements IsWholly Unreliable................................................................................................. 28

II. Dr. Maraiss Opinions Are Irrelevant and Needlessly Cumulative .................................. 32

III. Professor Blacks Opinions Should Be Precluded............................................................ 34

A. Professor Blacks Methods Are Not Reliable and Are Not AppliedReliably to the Facts of this Case.......................................................................... 34

B. Professor Black Is Not Qualified to Offer His Opinions ...................................... 38

C. Professor Blacks Proposed Rebuttal Testimony Impermissibly AddressesIssues Not Addressed By Defendants Initial Experts Reports........................... 40

i



4/61

IV. Professor Kraakmans Opinions Should Be Precluded .................................................... 42

A. Professor Kraakmans Opinions on Questions of Law Are Impermissible.......... 43

B. Professor Kraakman Is Not Qualified to Testify on the RemainingQuestions, He Unreliably Applies Unreliable Methods to Answer Themand His Opinions on These Questions Lack the Requisite Relevance ................. 44

C. Professor Kraakmans Proposed Rebuttal Testimony ImpermissiblyAddresses Issues Not Addressed By Defendants Initial Experts Reports.......... 49

CONCLUSION............................................................................................................................. 50

ii



5/61


6/61

Kumho Tire Co. v. Carmichael, 526 U.S. 137 (1999) .................................................................... 5

McDonough v. City of Quincy, 452 F.3d 8 (1st Cir. 2006).......................................................... 33

Nieves-Villanueva v. Soto-Rivera, 133 F.3d 92 (1st Cir. 1997)................................................... 43

North v. Cummings, No. 06-CV-174-D, 2007 WL 4616282 (D. Wyo. Mar. 27, 2007) .............. 40

Peak Interests, L.L.C. v. Tara Hills Villas, Inc., No. 8:06-CV-747, 2008 WL 650301(D. Neb. Mar. 5, 2008)............................................................................................................ 40

Pelletier v. Main St. Textiles, LP, 470 F.3d 48 (1st Cir. 2006) ................................................ 6, 43

Poulis-Minott v. Smith, 388 F.3d 354 (1st Cir. 2004).................................................................. 38

Ruiz-Troche v. Pepsi Cola of P.R. Bottling Co., 161 F.3d 77 (1st Cir. 1998) ..................... 5, 7, 32

Teachers Ret. Sys. of La. v. Hunter, 477 F.3d 162 (4th Cir. 2007)....................................... 11, 26

Seahorse Marine Supplies, Inc. v. P.R. Sun Oil Co., 295 F.3d 68 (1st Cir. 2002) ......................... 5

STS Software Sys., Ltd. v. Witness Sys., Inc., No. 1:04-CV-2111 (RWS),2008 WL 660325 (N.D. Ga. Mar. 6, 2008)............................................................................. 40

United States v. Buchanan, 964 F. Supp. 533 (D. Mass. 1997).................................................... 43

United States v. Diaz, 300 F.3d 66, 73 (1st Cir. 2002)................................................................... 5

United States v. Green, 405 F. Supp. 2d 104 (D. Mass. 2005)..................................................... 24

United States v. Mooney, 315 F.3d 54 (1st Cir. 2002). .................................................................. 6

Statutes and Rules

Fed. R. Evid. 402 ...................................................................................................................... 5, 32

Fed. R. Evid. 403 ............................................................................................................ 5, 6, 33, 49

Fed. R. Evid. 702 ................................................................................................................... passim

Fed. R. Civ. P. 26(a)(2)(B) ........................................................................................................... 40

Fed. R. Civ. P. 26(a)(2)(C) ........................................................................................................... 40

Fed. R. Civ. P. 37(c)(1)................................................................................................................. 40

iv



7/61

v

Other Authorities

Nihat Aktas et al., Event Studies with a Contaminated Estimation Period, 13 J. Corp. Fin.129 (2007)............................................................................................................................... 24

John Y. Campbell et al., Econometrics of Financial Markets (1997)............................... 13, 15, 23

Imre Karafiath, Using Dummy Variables in The Event Methodology, 23 Fin. Rev. 351(1988)...................................................................................................................................... 23

A. Craig MacKinley, Event Studies in Economics & Finance, 35 J. Econ. Literature 13(1997)...................................................................................................................................... 13

David I. Tabak & Frederick C. Dunbar, Materiality & Magnitude: Event Studies in theCourtroom, inLitigation Services Handbook: The Role of the Financial Expert(Roman I. Weil et al. eds., 3d ed. 2001) ..................................................................... 13, 15, 23



8/61

DEFENDANTS MEMORANDUM IN SUPPORT OF THEIR MOTION TO PRECLUDE

THE EXPERT OPINIONS OF SCOTT D. HAKALA, M. LAURENTIUS MARAIS,

BERNARD BLACK AND REINIER KRAAKMAN

Defendants Credit Suisse Securities (USA) LLC (CSFB), Credit Suisse (USA), Inc.

(Credit Suisse (USA)), Frank Quattrone, Elliott Rogers, Jamie Kiggen and Laura Martin

(collectively, Defendants) respectfully submit this memorandum in support of their motion

pursuant to Rule 702 of the Federal Rules of Evidence to preclude the expert opinions of Dr.

Scott D. Hakala, Dr. M. Laurentius Marais, Professor Bernard Black and Professor Reinier

Kraakman, the proposed expert witnesses of Class Representative Bricklayers and Trowel Trades

International Pension Fund (Plaintiff).

PRELIMINARY STATEMENT

At class certification, Defendants demonstrated that the opinions of Dr. Hakala,

Plaintiffs expert on materiality, loss causation and damages, were thoroughly unreliable. This is

because his event study the statistical technique on which he purported to base those opinions

did not conform to the established norms of financial economics.1 Defendants argued that Dr.

Hakalas opinions in support of class certification should be rejected for exactly the same reasons

that Judge Zobel of this Court precluded Dr. Hakala from testifying in In re Xcelera.com

Securities Litigation (Xcelera), Civ. A. No. 00-11649 (RWZ), 2008 U.S. Dist. LEXIS 77807

(D. Mass. Apr. 25, 2008), and for the same reasons that several other judges have recently

rejected Dr. Hakala as an expert. Although this Court did not rule on the admissibility of Dr.

Hakalas testimony at that time, it nonetheless acknowledged the seriousness of [D]efendants

challenges to [Dr. Hakalas] conclusions . . . as to market impact [and] loss causation, based on

1 (See, e.g., Defs. Class Cert. Oppn Br. [Dkt. # 152] 2, 27-30; Defs. Class Cert. Sur-Reply [Dkt. # 173]1-17; Defs. Apr. 15, 2008 Letter Br. [Dkt. # 225] 4-6; Defs. Notice of Supp. Auth. [Dkt. # 228] 1-2; Defs. DaubertMem. [Dkt. # 232] 1-12; Defs. Daubert Sur-Reply [Dkt. # 239] 1-5.)



9/61

both the legal framework and methodology that Dr. Hakala had employed. In re Credit

SuisseAOL Sec. Litig. (Credit SuisseAOL II), 253 F.R.D. 17, 30 n.16 (D. Mass. 2008).

After the close of fact discovery, Plaintiff initially served only one expert report to

support its case on the merits: yet another report by Dr. Hakala, complete with a new event

study on AOL Time Warner, Inc. (AOL) different from the one submitted in support of class

certification.2 Plaintiff now offers this report as its only evidence of market impact and loss

causation in opposition to Defendants Memorandum in Support of Their Motion for Summary

Judgment. (See, e.g., Pl.s S.J. Opp. [Dkt. # 288] 31-33, 54-65.) However, Dr. Hakala has done

nothing to remedy the numerous, glaring and fundamental defects in his opinions that

Defendants identified last year in their class certification briefing. To the contrary, his most

recent expert and rebuttal reports compound his earlier errors the very same ones that led to his

exclusion in Xcelera. For example:

Dr. Hakala continues to fail to control for confounding factors, and insteadspeculates, based on his event study, about the causes of AOLs stock pricemovements on days when there were numerous, contemporaneous disclosures, even

though he admits that the established techniques of financial economics provide nobasis for him to do so;

he continues to disregard the fundamental concept of market efficiency by analyzingreiterations of previously disclosed information as new corrective disclosures, eventhough in an efficient market, such as the one for AOL stock, these later disclosuresshould have no incremental effect on share price;

he continues to use a poorly defined methodology, which allows him to manufactureresults that cannot be replicated by other economists;

he continues to improperly overuse dummy variables, leading him to determine thatnumerous days throughout the period from January 12, 2001 through July 24, 2002

2 CSFB and Credit Suisse (USA), in turn, submitted expert reports from Ren M. Stulz, a financialeconomist, and John Deighton, a professor of marketing and internet advertising. (See Gesser Decl. Ex. 15(Corrected Expert Report of Ren M. Stulz, dated July 8, 2008 (Stulz Rpt.)); Gesser Decl. Ex. 17 (Expert Reportof John Deighton, dated May 1, 2008 (Deighton Rpt.)).)

2



10/61

(the Class Period)3 are statistically significant event dates when, in fact, theywould not be if Dr. Hakala had done his event study properly; and

he continues to base many of his opinions on sheer speculation, even when hisspeculation is plainly contradicted by the facts established through the voluminous

documentary and testimonial record in this case.

In addition, Dr. Hakala cannot demonstrate through the accepted techniques of financial

economics that CSFBs research reports had any measurable effect on AOLs stock price. That

is not a new difficulty for him. What is new, however, is that Dr. Hakala now attempts to

quantify the nonexistent market impact of Defendants reports based on a proxy that he

constructs from the movements of AOLs share price in response to public statements by other

analysts and by AOL itself regarding information that CSFB is never alleged to have known.

There is no support in logic much less in the academic literature for this approach.

Apparently recognizing Dr. Hakalas inability to remedy the flaws in his own report, or to

respond to the devastating criticisms of his methodology by CSFBs experts, Plaintiff tries to

redeem Dr. Hakala and address the clear deficiencies in his testimony through the rebuttal

reports submitted by three new experts: Dr. Marais and Professors Black and Kraakman. But

each of these rebuttal reports is just as flawed as Dr. Hakalas report. Dr. Marais whose

opinions are confined to rehabilitating Dr. Hakalas use of dummy variables cannot save Dr.

Hakala from exclusion because Dr. Marais did not actually test Dr. Hakalas methodology or

validate his results, nor did Dr. Marais address any of Dr. Hakalas numerous other

methodological faults.

Professors Black and Kraakman not only fail to rebut any of Defendants expert evidence

3 Although the certified Class Period ends on July 24, 2002, Dr. Hakala includes July 25, 2002 in his eventstudy to measure the effects of disclosures that he claims were made on July 24 after the close of the market.(Gesser Decl. Ex. 4 (Expert Report of Scott D. Hakala, dated Mar. 4, 2008 (Hakala Rpt.) 6 & n.4.) Thus, forconvenience only, Defendants references to Class Period in this memorandum include July 25, 2002.

3



11/61

(indeed, Professor Black admitted that he agreed with Professor Deighton), but their reports

suffer from fundamental problems of their own. For example, in opining that Defendants AOL

reports were both material and unreasonable, Professor Black failed to even consider the most

basic measure of reasonableness what other analysts covering AOL were saying even though

he himself admitted that such information would have been important to his opinion. Moreover,

at his deposition, Professor Black was forced to cross out line after line of his expert report and

withdraw his opinion that Defendants AOL reports were unreasonably optimistic, because, as he

candidly admitted, he had improperly based that opinion on data that did not exist at the time

Defendants issued their analyst reports. Finally, Professor Kraakman who is not a financial

economist essentially disqualified himself as an expert by admitting that he is probably no

more of an expert on the academic literature on which he was opining than anyone who has read

the same handful of articles that he has read (which, incidentally, he read for the first time

shortly before issuing his report in this case). And his principal opinion that the fraud-on-the-

market presumption of reliance can apply to research analyst statements addresses a legal issue

that this Court has already ruled on, not a fact issue for the jury.

For these reasons, the opinions of all four of Plaintiffs experts must be excluded.

ARGUMENT

As demonstrated in the materials submitted in support of Defendants Motion for

Summary Judgment, there is no evidence in the factual record to support a finding of loss

causation, reliance or materiality in this case. (See generally Defs. S.J. Mem. [Dkt. # 279], as

corrected; Defs. 56.1 [Dkt. # 280], as corrected; Defs. Reply S.J. Mem. [Dkt. # 296].) Unable

to establish these elements and thus unable to satisfy its burden of proof (whether in response to

Defendants summary judgment motion or at trial), Plaintiffs only recourse for salvaging its

case is to depend upon the result-driven, internally inconsistent and scientifically unsound

4



12/61

testimony of Dr. Hakala and the equally problematic testimony of three other experts in a bid to

show how CSFBs statements supposedly could have caused Plaintiffs alleged losses. Dr.

Hakalas testimony and that of Dr. Marais, Professor Black and Professor Kraakman fall far

short of the admissibility standards for expert evidence in myriad ways and must be rejected.

The testimony of a proposed expert may not be relied upon at summary judgment if it is

inadmissible. See, e.g., In re Williams Sec. Litig.WCG Subclass, ___ F.3d ___, No. 07-5119,

2009 WL 388048, at *1 (10th Cir. Feb. 18, 2009) (affirming exclusion of expert because his

theories of loss causation could not distinguish between the alleged fraud and confounding

factors, and affirming summary judgment for defendants); Hartford Ins. Co. v. Gen. Elec. Co.,

526 F. Supp. 2d 250, 254 (D.R.I. 2007) (granting motion to preclude expert and for summary

judgment). To be admissible, expert testimony must be both (1) relevant and (2) reliable.

Daubert v. Merrell Dow Pharms., Inc., 509 U.S. 579, 597 (1993); United States v. Diaz, 300 F.3d

66, 73 (1st Cir. 2002).4

The First Circuit has recognized that, [t]o be admissible, expert testimony must be

relevant not only in the sense that all evidence must be relevant, see Fed. R. Evid. 402, but also

in the incremental sense that the experts proposed opinion, if admitted, likely would assist the

trier of fact to understand or determine a fact in issue. Ruiz-Troche v. Pepsi Cola of P.R.

Bottling Co., 161 F.3d 77, 81 (1st Cir. 1998) (citing Daubert, 509 U.S. at 591-92); see also Fed.

R. Evid. 702 (expert testimony is permitted only [i]f scientific, technical, or other specialized

knowledge will assist the trier of fact to understand the evidence or to determine a fact in issue).

4In Daubert, the Supreme Court held that Federal Rule of Evidence 702 imposes a special obligation upona trial judge to ensure that any and all scientific testimony . . . is not only relevant, but reliable. Kumho Tire Co.v. Carmichael, 526 U.S. 137, 147-49 (1999) (quoting Daubert, 509 U.S. at 589). In Kumho, the Court held that atrial courts gatekeeping obligation under Daubert and Rule 702 applies to all expert testimony, not just toscientific expert testimony. Kumho, 526 U.S. at 147-49; see also Seahorse Marine Supplies, Inc. v. P.R. Sun OilCo., 295 F.3d 68, 81 (1st Cir. 2002); Diaz, 300 F.3d at 73 n.4.

5



13/61

Importantly, however, even relevant expert testimony may be excluded if its probative value is

substantially outweighed by the danger of unfair prejudice, confusion of the issues, or misleading

the jury, or by considerations of undue delay, waste of time, or needless presentation of

cumulative evidence. Fed. R. Evid. 403.

A proposed experts testimony must be not only relevant, but reliable as well. To be

reliable, an expert witness must be qualified . . . by knowledge, skill, experience, training, or

education. Fed. R. Evid. 702. Further, the expert testimony must be (1) based upon sufficient

facts or data and (2) the product of reliable principles and methods that the expert has

(3) applied . . . reliably to the facts of the case. Id. With respect to this requirement, the

Supreme Court has explained that a court need not admit an expert opinion that is inadequately

linked to the facts, and may exclude opinion evidence that is connected to existing data only by

the ipse dixit of the expert. A court may conclude that there is simply too great an analytical gap

between the data and the opinion proffered. Gen. Elec. Co. v. Joiner, 522 U.S. 136, 146 (1997);

see also Pelletier v. Main St. Textiles, LP, 470 F.3d 48, 55-56 (1st Cir. 2006) (affirming

exclusion of expert who failed to adequately base testimony on facts of the case). Daubert

enumerated four non-exclusive factors for courts to consider when determining whether expert

testimony is sufficiently reliable to be admitted: (1) whether the proffered theory or technique

can be and has been tested; (2) whether the proffered theory or technique has been subjected to

peer review and publication; (3) a proffered techniques known or potential rate of error and the

existence and maintenance of standards controlling the techniques operation; and (4) the level of

the proffered theory or techniques acceptance within the relevant community. Daubert, 509

U.S. at 593-94; see also United States v. Mooney, 315 F.3d 54, 62 (1st Cir. 2002).

The burden of showing that an experts testimony meets the standards for admissibility is

6



14/61

on the party who proffers that testimony. See Ruiz-Troche, 161 F.3d at 85. For the reasons set

forth below, Plaintiff has not met and cannot meet that burden here as to any of its purported

experts.

I. Dr. Hakalas Opinions Are Entirely Unreliable and Must Be PrecludedIn the last five years, courts around this country have repeatedly rejected the opinions of

Plaintiffs lead expert, Dr. Hakala, because the event studies he performs are not based on a

scientifically acceptable methodology and because his opinions are not the result of a

methodology that is reliably applied to the facts. See, e.g., Greenberg v. Crossroads Sys., Inc.,

364 F.3d 657, 666-67 (5th Cir. 2004) (declining to accept Dr. Hakalas opinions on reliance and

loss causation because he failed to control for confounding events on key corrective disclosure

day); Fener v. Belo Corp., 560 F. Supp. 2d 502, 506-07 (S.D. Tex. 2008) (rejecting Dr. Hakalas

opinions on loss causation because he failed to control for confounding factors and because he

fail[ed] . . . to substantiate [his] claim[s]); In re Omnicom Group, Inc., Sec. Litig., 541 F. Supp.

2d 546, 554 (S.D.N.Y. 2008) (rejecting Dr. Hakalas opinions on loss causation because his

event study at best incorrectly identifies corrective disclosures and fails to control for

confounding events); cf. In re Broadcom Corp. Sec. Litig., No. SACV 01275 GLTMLGX, 2005

WL 1403756, at *3 (C.D. Cal. June 3, 2005) (excluding Dr. Hakalas opinion on aggregate

damages under Daubert because his model was unreliable). Recently, Judge Zobel precluded Dr.

Hakala from testifying under Daubert and Federal Rule of Evidence 702 in Xcelera because his

event study did not conform to the norms of financial economics in (1) its failure to control for

confounding factors, (2) its disregard of market efficiency, as demonstrated by Dr. Hakalas

ignoring the effects of prior disclosures, (3) its failure to analyze relevant event dates and (4) its

use of dummy variables. 2008 U.S. Dist. LEXIS 77807, at *2-7.

For these very same reasons, Dr. Hakalas event study his third on AOL and his second

7



15/61

in this case is equally faulty.5 Moreover, Dr. Hakalas improper use of a speculative proxy to

quantify the nonexistent market impact of CSFBs AOL reports provides an independent basis

for precluding his testimony here.6

A. Dr. Hakalas Event Study Fails to Control for Confounding Events andContravenes Market Efficiency

1. Dr. Hakalas Failure to Control for Confounding InformationRenders His Event Study Unreliable

As Defendants have shown in past briefing, Dr. Hakalas event studies include allegedly

inflationary and corrective disclosure dates that are confounded by the publication of information

unrelated to the supposed disclosures on which Plaintiff has focused. (See Defs. S.J. Mem. 63-

65; Defs. Daubert Mem. 11-12; see also Stulz Rpt. 5(e).) This rendered Dr. Hakalas

conclusions at class certification thoroughly unreliable, because there was no way for him to use

his event study, consistent with accepted methods of financial economics, to determine which (if

any) of two or more pieces of confounding news actually caused AOLs stock price to change.

(See, e.g., Defs. Daubert Mem. 11-12.) In his latest event study, Dr. Hakala opines that over the

course of the Class Period, disclosures related to the alleged fraud occurred on 57 separate dates,

on which he claims that news released to the market either artificially inflated AOLs share price

(21 inflationary dates) or removed artificial inflation from AOLs share price by partially

revealing the fraud and purportedly causing damages (36 corrective disclosure dates), which

Dr. Hakala attributes to Defendants. Most of these dates, however, are confounded. Dr.

Hakalas continued failure to control for confounding events thus renders Dr. Hakalas current

5 What event studies are and the manner in which they are performed is discussed at Part I.B infra.

6 In his March 2008 report, Dr. Hakala has provided, for information purposes, his estimates ofaggregate damages, although he has indicated that Plaintiff may choose not to present this information at trial.(Hakala Rpt. 16.) Defendants reserve the right to move to exclude such testimony if and when Plaintiff chooses tointroduce it.

8



16/61

report methodologically unreliable and requires its exclusion on Daubert grounds. Xcelera, 2008

U.S. Dist. LEXIS 77807, at *6-7; see also In re Williams, 2009 WL 388048, at *10.

Defendants have established an ample record documenting the fact that Dr. Hakalas key

alleged corrective disclosure dates were confounded by unrelated information. For instance:

February 1, 2001: Information released to the market included (i) a positiveCSFB report, (ii) several analyst reports containing material information aboutAOL in addition to that allegedly concealed by Defendants, (iii) disappointingAOL earnings and (iv) media commentary on AOL. (56.1 204.)

July 18, 2001: AOL released disappointing earnings that contemporaneous newsreports attributed to several factors, only some of which related to allegations inthe Complaint. (See Gesser Decl. Ex. 22 (AOL Time Warner Reports Second

Quarter EBITDA of $2.5 Billion, Business Wire (July 18, 2001)).)

July 19, 2001 and February 5, 2002: Multiple analysts released reports on AOL(see Hakala Report Ex. B-1) and Plaintiff fails to identify any method fordetermining which of the several disclosures, if any, actually caused the stockprice to move (see Gesser Decl. Ex. 6 (Hakala 2007 Tr.) 70:25-71:3; Defs. S.J.Mem. 64).

February 20, 2002: Lehman Brothers released a report downgrading AOL as aresult of four factors, only one of which pertained to Plaintiffs allegations (i.e.,AOLs exposure to the weak advertising market) (56.1 283; see also Gesser

Decl. Ex. 13 (Kraakman Tr.) 172:5-173:6), and the information on which theanalyst based her negative views regarding AOLs advertising prospects had beenpreviously disclosed by the same analyst, by AOL and by the press, without anystatistically significant effects on AOLs stock price. (Defs. S.J. Mem. 64-65;Defs. Class Cert. Sur-Reply 17-18 n.7; Defs. Letter Br. 4-5.)

Similarly, Dr. Hakalas key inflationary dates such as September 19, 2001 were clearly

confounded by reports from other analysts, for which Dr. Hakala makes no attempt to control

because no accepted method for doing so exists. (See Defs. S.J. Mem. 70; Stulz Rpt. 94.)

Indeed, although Dr. Hakala has admitted that there is no peer reviewed way to deal

with confounding events[,] period[,] in the literature (Hakala 2007 Tr. 70:25-71:3), that has not

stopped him from opining that Defendants are responsible for numerous increases and decreases

in AOLs share price on days that are thoroughly confounded by unrelated news from other

9



17/61

analysts and from AOL itself. Dr. Hakala made no attempt at class certification to disentangle

the many factors affecting AOLs stock price on such disclosure days in his event study. (See

Hakala 2007 Tr. 70:22-71:8, 211:7-18.) In his latest event study, he similarly conducted no

analysis to disaggregate the effects of confounding disclosures; instead, he simply drew

inferences about which of several pieces of confounding news affected AOLs share price based

on his subjective and untestable feelings about whether the news was significantly positive or

negative, and the direction in which AOLs stock price moved that day. (See, e.g., Gesser Decl.

Ex. 7 (Hakala 2008 Tr.) 84:9-85:17 (explaining his general approach); id. 394:7-12 (Q. So

you dont view [Oct. 17, 2001] as confounded? A. It is partially confounded but I dont view it

as confounded enough to ignore the effect of the analyst on that day on the stock price.).) Dr.

Hakalas instinct for when it is appropriate to ignore confounding news in assessing causation is

obviously not a reproducible scientific methodology that is supported by the academic finance

literature.

Therefore, Dr. Hakalas conclusion that AOLs price movements on numerous days

throughout the Class Period were caused by Defendants and not caused by any number of other

speakers unrelated statements about AOL is pure speculation that must be excluded. See

Xcelera, 2008 U.S. Dist. LEXIS 77807, at *6-7; In re Williams, 2009 WL 388048, at *10

(stating that experts failure to show why the February 4 losses should be attributed to the

revelation of the fraud and not other non-fraud related news renders his methodology

unreliable); cf. Dura v. Broudo, 544 U.S. 336, 343 (2005) (describing the tangle of factors

affecting market price and emphasizing that, while an inflated purchase price may suggest that a

misrepresentation touches upon a subsequent economic loss, [t]o touch upon a loss is not

to cause a loss, and it is the latter that the law requires (citation omitted)).

10



18/61

2. Dr. Hakalas Disregard of Prior Disclosures ContradictsMarket Efficiency and Renders His Event Study Unreliable

Dr. Hakalas current report is further flawed because it includes inflationary and

corrective event days on which the information disclosed to the market was not new. Dr.

Hakalas report thus contravenes this Courts finding and his own opinion that AOL stock

traded in an efficient market. See Credit SuisseAOL II, 253 F.R.D. at 32; (Hakala Rpt. 19).

In an open and efficient market, a companys stock reacts immediately to new information that is

important to reasonable investors. In re PolyMedica Corp. Sec. Litig., 432 F.3d 1, 14 (1st Cir.

2005). Conversely, an efficient market should not react at all to reiterations of previously

released information. Teachers Ret. Sys. of La. v. Hunter, 477 F.3d 162, 187 (4th Cir. 2007).

As Defendants have shown, Dr. Hakalas event studies in this case disregard market efficiency

for exactly the same reasons that led to the exclusion of his testimony by Judge Zobel in Xcelera,

and his opinions here should similarly be excluded. (See, e.g., Defs. Daubert Mem. 9-10; Defs.

Class Cert. Sur-Reply 8-9, 16 n.6, 17-18 n.7; Defs. Letter Br. 4-5; Defs. S.J. Mem. 61-62, 69-

71.)

Judge Zobel found that Dr. Hakalas event study in Xcelera could not be squared with

the theory of market efficiency because (1) his study found that it took multiple days for

Xceleras stock price to incorporate allegedly material information and (2) the same information

had to be disclosed several times before Dr. Hakalas event study showed that it was

incorporated into the stock price. Xcelera, 2008 U.S. Dist. LEXIS 77807, at *4-6. One or both

of these flaws can be found with respect to four out of the five key event days that Dr. Hakala

claimed at class certification both were curative disclosures and had a statistically significant

price effect on AOLs stock (two of which August 14, 2001 and July 24, 2002 are no longer

11



19/61

even relevant to his analysis).7 The same flaws apply to numerous other purported inflationary

and corrective disclosure dates throughout the Class Period that Dr. Hakala includes in his

present report. (See Defs. Daubert Mem. 10; Defs. S.J. Mem. 61-62, 64-65.)

For instance, as noted above (supra Part I.A.1), the February 20, 2002 Lehman Brothers

report that Dr. Hakala views as a corrective disclosure repeated information about AOLs fourth

quarter 2001 intracompany advertising that the same analyst had disclosed, without any

statistically significant price effects, in a report on January 31, 2002. (Compare Gesser Decl. Ex.

24 (Lehman Bros. Report dated Feb. 20, 2002) with Gesser Decl. Ex. 23 (Lehman Bros. Report

dated Jan. 31, 2002); see also Stulz Rpt. 45, 62; Defs. Daubert Mem. 10; Defs. S.J. Mem.

64.) Similarly, Dr. Hakala maintains that February 5, 2001 was a very important inflationary

date due to a CSFB report issued that day, yet he admitted at his deposition that the report did not

contain any new information. (Hakala 2008 Tr. 206:3-24, 210:13-214:18 ([S]ometimes a report

that might not change anything . . . can sometimes have a positive effect. And thats an example

of whats happening here.).) Such inconsistencies in approach and methodology require

exclusion of Dr. Hakalas testimony.

B. Dr. Hakalas Event Study Methodology Is Fundamentally FlawedBeyond Dr. Hakalas failure to control for confounding events and his disregard of

market efficiency, his latest event study suffers from even more fundamental flaws because of

the idiosyncratic and unreliable methodology he used to design and perform the study.

To understand Dr. Hakalas numerous and significant deviations from the accepted

methodology of financial economics, it is first necessary to review how a standard event study is

7 The key event days that Dr. Hakala identified at class certification were 7/18/01, 8/14/01, 2/20/02,7/24/02 and 7/25/02. (See Pl. 5/09/08 Daubert Mem. [Dkt. # 229] 4.) As noted above, August 14, 2001 and July 24,2002 are no longer relevant to Dr. Hakalas analysis. (See Hakala Rpt. Ex. C-1a, at 5 & Ex. C-2a, at 2 (showingzero relevant events for each day).)

12



20/61

performed. An event study measures the impact of a specific event on the value of a firm.

(Gesser Decl. Ex. 14 (Declaration of Ren M. Stulz, dated Apr. 26, 2007 (Stulz Decl.)) 10

(quoting A. Craig MacKinley, Event Studies in Economics & Finance, 35 J. Econ. Literature 13,

13 (1997)).)

Most event studies have two basic steps.8

First, an economist performs a regression to

estimate the relationship between a stocks actual return (e.g., the difference between closing

prices on two consecutive days) and the movement of one or more indices representing an

average of the stock prices for several companies that make up the market and/or industries in

which the firm operates. This step allows the economist to predict how the stock should move

on any given day based on the movement of the indices (an expected return), and thereby

controls for factors that affect all companies within a particular market or industry.9 Second, the

expected return estimated through the regression is used as the baseline against which the stocks

actual return on preselected event days is measured. The expected return is thus a measured

expectation of what the normal stock price movement would be if the event had not occurred. If

the difference between the expected return and actual return (known as the abnormal return) on

8 The description of event study methodology that follows is drawn from Stulz Decl. 9-13; Gesser Decl.Ex. 18 (John Y. Campbell et al., Econometrics of Financial Markets, 150-52 (1997)); and Gesser Decl. Ex. 19(David I. Tabak & Frederick C. Dunbar, Materiality & Magnitude: Event Studies in the Courtroom, inLitigationServices Handbook: The Role of the Financial Expert 19.5-19.6 (Roman I. Weil et al. eds., 3d ed. 2001)). (Seegenerally Defs. Daubert Mem. 3-5.)

9 An economist must exclude from the regression the possible effects of the preselected events for which heor she is testing. Otherwise, the predicted return will already include such effects, if any, which may cause the

predicted return to be higher or lower than it would otherwise be if the event had been excluded. Suchcontamination could affect the findings in the second step of the event study. To avoid this problem, economiststypically perform the regression for a period that precedes the event subject to analysis. (Stulz Decl. 12); seealso Campbell et al. supra note 8, at 152. In this case, such an approach was not feasible because the first day of theClass Period was also the first day that AOL traded as a combined company. (Stulz Decl. 12.) In thesecircumstances, the economist may perform the regression over a period that covers the events that were preselectedfor testing, but will exclude the event dates themselves from the regression (either by using dummy variables orotherwise), as Professor Stulz did. (See id.; see generally Defs. Daubert Mem. 4-5.) It is not, however, permissibleto use dummy variables for all manner of non-event days, as Dr. Hakala does. (See infra Part I.B.2(a).)

13



21/61

an event day is sufficiently large, then it may be attributed to the event occurring on that day,

provided that there are no confounding events that same day. The magnitude of the abnormal

return is assessed through statistical significance, which is determined by comparing the

abnormal return on the event day to the standard deviation of all abnormal returns during the

period over which the regression was estimated (the estimation window).10

Dr. Hakalas event study diverges in important respects from this procedure. Among

other things, Dr. Hakala fails to properly preselect relevant events to test, and adds an entirely

new step to his regression, through which he excludes numerous, nonevent days (which he

confusingly calls material events) from his estimation period by using dummy variables.

1. Dr. Hakala Has Failed to Use Relevant Event Dates and HasManufactured the Relevance of the Events He Supposedly Tested

The event dates that Dr. Hakala analyzed in the event study attached to his expert report

are largely irrelevant to the facts of this case, just as they were in his class certification

declaration and in his Xcelera report.11

In Xcelera, Judge Zobel rejected Dr. Hakalas testimony

because, [q]uite simply, his theory does not match the facts. Xcelera, 2008 U.S. Dist. LEXIS

77807, at *4. In that case, Dr. Hakala had mischaracterized a single key date as relevant

despite the lack of any evidence that the disclosures on that date actually corrected the

10 For example, if an event is statistically significant at the 95% confidence level (t-statistic=1.96), it meansthat an event of the same or greater magnitude should randomly occur on only 5% of days. (Stulz Decl. 8 n.5.)

Dr. Hakala uses the 90% confidence level in his event study, which is less accepted in the academic literature.(Hakala Rpt. 34; see also Stulz Decl. 13; Gesser Decl. Ex. 16 (Stulz 2007 Tr.) 170:6-14.) Dr. Hakalas lowerthreshold for significance is another factor predisposing his analysis to error.

11 As Defendants demonstrated in their summary judgment brief, all of Dr. Hakalas event days areirrelevant because there is no evidence in the record linking the alleged corrective disclosures regarding advertisingmarket risk, layoffs or accounting investigations to anything Defendants allegedly knew but did not disclose in theirAOL analyst reports. (Defs. S.J. Mem. 65-68.) Defendants will not repeat those arguments here, and instead focuson other incurable problems in Dr. Hakalas relevance determinations.

14



22/61

defendants alleged fraud. Id. at *3-*4. Here, Dr. Hakala lists 57 relevant days in Exhibits C-

1a and C-2a of his expert report, and 118 relevant days in Exhibit B-1, most of whose

relevance he either cannot explain at all, or can explain only through pure ipse dixit, which

cannot be credited. See, e.g., Hakala 2008 Tr. 304:24-305:4 (explaining that February 8, 2001

Merrill Lynch report was relevant because Blodget would have had a different impact had

Credit Suisse made a different statement); cf. Joiner, 522 U.S. at 146 (courts may exclude

opinion evidence which is connected to existing data only by the ipse dixit of the expert). Dr.

Hakalas unscientific and arbitrary relevance determinations, which have little to do with the

allegations or facts in this case and make no sense even under his own definition of relevance,

are merely after-the-fact rationalizations for Plaintiffs recovery that do not follow any accepted

methodology of financial economics.

In his latest expert report, Dr. Hakala never really defines the events he is testing. At the

beginning of an event study, an economist must identify the events whose stock-price impact he

intends to test. Campbell et al., supra note 8, at 151; Tabak & Dunbar, supra note 8, at 19.4.

Failure to rigorously identify the events to be tested creates a risk that the relevant events will be

determined in an arbitrary fashion after the event study has been performed, making it

impossible for a different economist to replicate the analysis. In securities fraud cases, the

events of interest usually include all the alleged disclosures of fraud, the dates of the fraudulent

statements, or both. Tabak & Dunbar, supra note 8, at 19.4. In this case, one would thus expect

Dr. Hakala to select as his relevant events the dates of Defendants AOL analyst reports and the

dates alleged as corrective disclosures in the Complaint. That, however, is not what Dr. Hakala

does. Rather, he engages in after-the-fact speculation exactly what the proper event study

methodology is designed to prevent.

15



23/61

Instead of preselecting events (as that term is normally used in the financial literature)

before performing his event study, Dr. Hakala identifies what he calls material event days

essentially, all AOL-specific news days, whether or not the news is in any way related to the

Complaints allegations. (Hakala Rpt. 30, 33 n.17.) As discussed below, Dr. Hakala then uses

dummy variables to exclude all of these material event days when he performs his regression.

These material events, however, are not the events that Dr. Hakala tests for significance in

the next step of his event study. Indeed, Dr. Hakala made no advance determination about which

events he needed to test to assess the effects of Defendants analyst reports and alleged

omissions on AOLs stock price.

Although Dr. Hakala purports to assess, after the fact, whether certain of his material

events were relevant to this case (see, e.g., Hakala Rpt. 44; id. at Exs. B-1, C-1a, C-2a), he

offers only vague and conflicting definitions of what it means for an event to be relevant.12

For

example, Dr. Hakalas full event study summary (Ex. B-1) inexplicably includes more than

twice as many relevant events (118) as are included in his event study summaries for

advertising-related claims and layoff/accounting-related claims (Exs. C-1a and C-2a,

respectively). With respect to the set of relevant events listed in Exhibits C-1a and C-2a, Dr.

Hakala states that [r]elevant events are those that either would not have occurred but for the

sustained allegations in the Complaint or would have occurred (or equivalent disclosure events

would have occurred) at or before the beginning of the Class Period but for the sustained

allegations in the Complaint. (Id. 44 (footnote omitted).) This definition, however, is utterly

meaningless, and cannot explain Dr. Hakalas event selection for at least two reasons.

12 Dr. Hakalas corrective disclosure dates can be identified in Exhibit C-1a if a date has both a RelevantEvents value other than zero and a negative 1 Day Effect value. Similarly, Dr. Hakalas inflationary dates aredates with both a Relevant Events value other than zero and a positive 1 Day Effect value.

16



24/61

First, as Defendants have repeatedly shown, many of Dr. Hakalas relevant events

concern matters (e.g., AOLs quarterly earnings releases) that had not happened at the start of the

Class Period and thus contrary to Dr. Hakalas definition of relevance could not, at that time,

have been known to or disclosed by CSFB or anyone else.13

This flaw is apparent in Dr.

Hakalas analysis of July 25, 2002 as a relevant disclosure date. (Hakala Rpt. Ex. C-2a, at 2.)

The new information that Dr. Hakala believes was disclosed that day that the SEC was

investigating AOL in response to articles the Washington Post had published the prior week

regarding AOLs accounting for certain advertising transactions simply did not exist at any

point in time when Defendants were issuing research on AOL. (See Defs. S.J. Mem. 11-15;

Defs. Daubert Mem. 8; Hakala 2007 Tr. 180:7-22; Defs. Class Cert. Sur-Reply 13-14.) Such a

disclosure obviously could not have been made at or before the beginning of the Class Period

eighteen months earlier, in January 2001.

Second, the relevance of many of Dr. Hakalas relevant event dates is directly

contradicted by the sustained allegations in the Complaint, and thus cannot be relevant

according to the definition that Dr. Hakala provides in his report. For example, on February 1,

2001, CSFB issued its first complete analyst report on AOL including the first issuance of the

$75 price target and 2001 revenue and EBITDA projections that Plaintiff claims were

misleading. (Second Consolidated Amended Complaint (the Complaint or SCAC) [Dkt.

# 92] 22.) On March 7, 2001, CSFB issued a report on the new AOL Networks division. The

Complaint alleges that this report was misleading because it did not include Laura Martins

purported belief that Networks was increasingly unlikely . . . [to] hit our P&L estimates.

13 To talk of equivalent disclosures being made for such events at or before the beginning of the ClassPeriod is equally nonsensical, especially in light of the fact that January 12, 2001, the first day of the Class Period,is the day that the America OnlineTime Warner merger closed.

17



25/61

(SCAC 26.) Both of these days are relevant for Dr. Hakala, but rather than treating them as

inflationary as the Complaint alleges, Dr. Hakala instead deems them to be corrective disclosures

i.e., revelations of the truth as to CSFB, which removed artificial inflation from AOLs stock

price, causing it to decline. (Hakala Rpt. Ex. C-1a, at 1, 2.) Dr. Hakalas definition of relevance

provides no explanation for this decision, which appears to be based on little more than the

decline in AOLs share price on those days and contemporaneous news reports unrelated to

CSFB. (See id.) Dr. Hakalas disregard of the Complaints allegations with respect to the

relevance of Defendants AOL reports is not isolated to just a few event dates. In fact, of the 35

CSFB reports on AOL that the Complaint alleges are misleading, Dr. Hakala deems 22 to be not

relevant to this case at all. (Compare Stulz Decl. Ex. 1 with Hakala Rpt. Ex. C-1a.)14

With respect to numerous other dates, Dr. Hakala makes no attempt to link specific

relevant events to specific alleged misstatements by CSFB or specific information that CSFB

allegedly knew and failed to disclose. It is thus impossible to discern how many of the relevant

events listed in Dr. Hakalas exhibits have anything to do with this case. For example:

February 8, 2001: According to Dr. Hakala, Henry Blodgett of Merrill Lynchinitiated research coverage of Microsoft with a downgrade but reiterated a positiverating for AOL. CSFB did not issue any report on AOL. (Hakala Rpt. Ex. C-1a, at2.) Rather than treat this day as unrelated to the case as he should have, Dr. Hakaladeems it a relevant inflationary event and increases the amount of damages heattributes to Defendants based on this event. (Id.)

April 4-5, 2001: According to Dr. Hakala, on April 4, 2001, Reuters reported on apossible deal between AOL and Dreamworks SKG, while Thomas Weisel Partnersmade positive comments about AOL. (Id. Ex. C-1a, at 3.) The following day, UBS

issued a positive report on AOL, Lehman Brothers issued a positive report on Yahoo!and Prudential cut its price target for AOL. (Id.) On neither day did CSFB issue areport on AOL. But instead of treating these events as entirely unrelated to this case,Dr. Hakala again deems them to be relevant inflationary events, and increases the

14 The CSFB reports that are not relevant according to Dr. Hakala are dated 1/16/01, 3/5/01, 3/8/01,3/21/01, 3/23/01, 4/10/01, 4/16/01, 5/15/01, 5/16/01, 5/22/01, 5/29/01, 6/4/01, 6/12/01, 6/25/01 (2 reports), 8/2/01,8/22/01, 9/25/01, 10/18/01, 11/26/01, 12/5/01 and 1/30/02.

18



26/61

damages he attributes to Defendants based on these events. (Id.)

June 7, 2001: According to Dr. Hakala, Deutsche Bank issued a positive reportstating that AOL CEO Gerald Levin is confident about meeting 2001 guidance andMerrill Lynch issued a report rating AOL a buy. (Id. Ex. C-1a, at 4.) Once again,

this date should be irrelevant. But even though CSFB made no statements aboutAOL and the disclosures appear to be positive, Dr. Hakala deems the event to be arelevant corrective disclosure, and assesses damages to Defendants as a result.(Id.)15

At his deposition, Dr. Hakala could not plausibly explain why most of CSFBs reports are

irrelevant to his analysis, while the seemingly unrelated reports of analysts at other firms are

relevant inflationary or deflationary events attributable to CSFB. Indeed, Dr. Hakala admitted

that if [he] marked something as relevant on an a priori ground [he] stuck with it as relevant

[there]after, regardless of what he found after studying the actual disclosures on the event day.

(Hakala 2008 Tr. 367:18-22.) Incredibly, when asked if he continued to do so [e]ven if [a

relevance determination] looked counterfactual after the fact, Dr. Hakala responded, Yeah,

yeah. (Id. 367:23-25.)

Dr. Hakalas relevance determinations in this case are thus even more problematic than

they were in Xcelera. See Xcelera, 2008 U.S. Dist. LEXIS 77807, at *3-*4 (excluding Dr.

Hakalas opinions because his event study focused on the wrong dates). As Defendants

demonstrated in prior briefing, Dr. Hakalas class certification declaration suffered from exactly

this same flaw (see Defs. Daubert Mem. 7-9; Defs. Class Cert. Sur-Reply 11-16), and it is now

obvious that Dr. Hakala is either unwilling or unable to limit his analysis to events that actually

15 Other event dates that Dr. Hakala finds relevant despite their obvious lack of connection to this caseinclude August 29, 2001, February 5, 2002, March 12, 2002 and April 18, 2002. Dr. Hakala treats all of these datesas corrective disclosures apparently due to statements by other analysts regarding advertising and AOLs financialprospects (id. Ex. C-1a, at 5, 8, 9), even though (1) CSFBs own estimates for AOLs financial prospects at the timeof these corrective disclosures were lower or less optimistic than those contained in the purported correctivedisclosures, (2) CSFB was no longer covering AOL at the time of the disclosures, (3) the disclosures containpositive news about AOL or (4) the disclosures concern issues CSFB is never alleged to have misstated (e.g.,subscriber growth and broadband issues). (See Defs. S.J. Mem. 60-61.)

19



27/61

are relevant to the allegations of this case and the facts established in discovery. Accordingly,

just as in Xcelera, his opinions must be excluded here as unreliable.

2. Dr. Hakala Continues to Improperly Use Dummy Variables, WhichRenders His Methodology Impossible to Replicate

(a) Dr. Hakalas Continued Overuse of Dummy VariablesRequires His Exclusion

Dr. Hakalas persistent, incorrect use of dummy variables to manufacture statistical

significance where there is none similarly requires his opinions to be excluded here, just as they

were in Xcelera.16 See 2008 U.S. Dist. LEXIS 77807, at *2-3. At class certification, Defendants

demonstrated that Dr. Hakalas use of dummy variables in his first event study in this case

deviate[d] significantly from the well-documented event study method in finance literature,

and seriously biased his results in favor of finding statistical significance, contrary to what a

standard event study would find.17 In his latest event study, Dr. Hakala abuses dummy variables

even more egregiously, using 79 (60%) more dummy variables during the Class Period than in

his earlier event study, and consequently finding 44 (68%) more days during the Class Period to

be statistically significant at the 90% confidence level. (Compare Hakala Rpt. Exs. C-1a & C-2a

with Hakala Decl. Ex. B; see also Stulz Rpt. 104 & Ex. 3.)18

Indeed, as Professor Stulz,

CSFBs financial economics expert, has found, the most remarkable aspect of Dr. Hakalas

approach is that it unfailingly inflates the statistical significance of abnormal stock returns of the

subject company. (Stulz Rpt. 100.)

16 A dummy variable is a variable that takes the values of 1 or 0 to indicate the presence or absence ofsome effect. (Stulz Decl. 26 n.27.) Dr. Hakala uses them to effectively exclude days on which there weredisclosures of material events i.e., AOL-specific news from his regression analysis. See supra Part I.B.1.

17 (Stulz Decl. 25-27; see also Defs. Class Cert. Sur-Reply 16 n.6; Defs. Daubert Mem. 3-7; Defs.Daubert Sur-Reply 3-5.)

18 In total, Dr. Hakala now finds 109 statistically significant events at the 90% confidence level (all withinthe Class Period), as compared to 81 (65 within the Class Period) at class certification.

20



28/61

Just as he did in Xcelera and at class certification, Dr. Hakala uses dummy variables in

his latest event study to control for what he calls material events effectively, for every date

on which he claims there was any news at all about [AOL] that might have affected the stock

price, Xcelera, 2008 U.S. Dist. LEXIS 77807, at *3. In the very same words he used to explain

what he did in Xcelera and at class certification, Dr. Hakalas latest expert report states that he

used these dummy variables to control for all days when potentially material information [about

AOL] came into the market. (Hakala Rpt. 30 (emphasis added); Gesser Decl. Ex. 3

(Declaration of Scott D. Hakala, dated Feb. 28, 2007 (Hakala Decl.)) 12 (verbatim); Gesser

Decl. Ex. 2 (Expert Report of Scott D. Hakala filed in In re Xcelera.com Sec. Litig., No. 00-CV-

11649 (RWZ), dated Apr. 26, 2007 (Xcelera Rpt.)) 31 (verbatim).)

Indeed, the only difference in Dr. Hakalas use of dummy variables in his current report

from his use of dummy variables in Xcelera and at class certification is the prodigious increase in

their number here. In Xcelera, the dummy variables exclude[d] over 130 of the 343 trading

days in the Class Period, or 38%. 2008 U.S. Dist. LEXIS at *3. At class certification, Dr.

Hakala used dummy variables to exclude 161 of 421 trading days (Hakala Decl. 15, 17), again

amounting to roughly 38% of his sample (Stulz Decl. 26). Now, however, Dr. Hakala uses

dummy variables to exclude 211 of 388 trading days more than half of the total trade days

considered in the analysis. (Hakala Rpt. 35 (referencing 216 events) (emphasis added); see

also Hakala Rpt. Ex. B-1 (identifying 217 events, of which 6 occur before the Class Period).)

In other words, in the extreme approach to dummying days out that Dr. Hakala takes in [his]

Report supposedly to control for aberrant observations created by AOL-specific news days, the

result is that a stock return is more likely to be aberrant than not. (Stulz Rpt. 100.)

Dr. Hakalas flawed approach essentially depresses his baseline measurement of the

21



29/61

volatility in AOLs stock price and invariably results in a greater number of days with larger

statistical significance. (Stulz Rpt. 100.) By using dummy variables to exclude material

events i.e., all days on which he identified a potentially material company-specific news

event (Hakala Rpt. 35) Dr. Hakala systematically removes from his regression those days

that tend to have larger stock-price reactions (Stulz Rpt. 100; see Stulz Decl. 26 (explaining

that any companys returns, including in this case AOLs, are likely to be more volatile on days

with news than on other days)). Thus, Dr. Hakala estimates AOLs expected returns based only

on the least volatile days the days when AOLs actual returns are likely to most nearly track

market/industry returns which leads to an understatement of what is [a] normal price

movement for AOL. (Stulz Rpt. 100; Stulz Decl. 26 (explaining that Dr. Hakalas use of

dummy variables has the effect of creating a downward bias in his estimate of the volatility of

abnormal returns).) Consequently, when Dr. Hakala calculates AOLs abnormal returns on the

relevant event days (when AOL-specific news does enter the market and returns are more

volatile) in the next step of his event study, those abnormal returns are far more likely to be large

and statistically significant than they would be if Dr. Hakala had used the accepted methodology

for event studies. (See Stulz Decl. 26; Defs. Daubert Mem. 5-6.)

This is not merely a theoretical problem with Dr. Hakalas report. At class certification,

Dr. Hakala had already found a number of supposedly relevant event dates to be significant

based on his improper use of dummy variables, including August 13, 2001, the date on which Dr.

Hakala asserts that the relevant truth about layoffs was disclosed to the market. (See Defs.

Daubert Mem. 6.) This event would not have been significant for Dr. Hakala had he performed

his event study in accordance with accepted methodology, as Professor Stulz did. (See Gesser

Decl. Ex. 25 (Stulz 2007 Tr. Ex. 2), at 3 (showing t-statistic of -1.13 for Aug. 13, 2001).) In Dr.

22



30/61

Hakalas latest event study, the 44 newly significant event dates which Dr. Hakala did not

consider statistically significant in 2007 but now does as a result of his using additional dummy

variables include still more relevant events that are critical for his analysis and Plaintiffs

case, such as February 5, 2001, the date CSFB issued a Media Week report on AOL, which is the

single largest inflationary event in Dr. Hakalas study. (Compare Hakala Decl. Ex. B, at 1 (t-stat

= 1.33 for 2/5/01) with Hakala Rpt. Ex. C-1a, at 1 (t-stat = 2.01 for 2/5/01).)

As Judge Zobel found in Xcelera, no peer-reviewed journal supports the view that

dummy variables may be used on all dates on which any company news appears, as Dr. Hakala

has done here. 2008 U.S. Dist. LEXIS 77807, at *3. Professor Stulz likewise know[s] of no

event study published in a peer-reviewed academic journal that uses [Dr. Hakalas] approach.

(Stulz Rpt. 99.) Indeed, Defendants have shown that academic texts that discuss the use of

dummy variables do not do so in the context of explaining how to perform event studies, and

caution that dummy variables should be used sparingly. (See Defs. Daubert Sur-Reply 4-5 &

n.6.) Articles that do explain how to perform event studies do not recommend the use of dummy

variables, at least not in the way Dr. Hakala has used them. See, e.g., Campbell et al., supra note

8, at 151-52 (no discussion of dummy variables); Tabak & Dunbar, supra note 8, at 19.4-19.6

(same); (Gesser Decl. Ex. 20 (Imre Karafiath, Using Dummy Variables in The Event

Methodology, 23 Fin. Rev. 351 (1988)) at 352 (no discussion of using dummy variables to

control for company-specific news days, but explaining that the same results as the traditional

[2-step] method for performing event studies may be obtained in one step by using dummy

variables for the events being tested (emphasis added)), cited in Hakala Rpt. 33 n.17). Even

Dr. Marais, whom Plaintiff has put forward as an expert in support of Dr. Hakalas methodology,

admits that there is no article in the academic literature that does exactly what Dr. Hakala did

23



31/61

[with dummy variables] in this case. (Gesser Decl. Ex. 9 (Marais Tr.) 105:2-4.)19

Accordingly, Dr. Hakalas dummy variable methodology, which is fundamental to his

analysis and has clearly biased his results in favor of finding statistical significance, cannot be

considered a generally accepted method of financial economics and his opinions must be

excluded here as a result.

(b) Dr. Hakalas Approach to Dummy Variables Cannot BeReplicated by Other Economists or Even Dr. Hakala

Dr. Hakalas dummy variables methodology is so arbitrary and mutable that his findings

cannot be replicated. Dr. Hakala has performed three event studies on AOL, each of which

purports to use a similar methodology and covers a similar time period, yet each of which

produces different results. A hallmark of a reliable expert methodology is that its results can be

replicated by a different analyst who uses the same methodology. See, e.g., United States v.

Green, 405 F. Supp. 2d 104, 120 (D. Mass. 2005) (Gertner, J.) (Reproducibility is an essential

component of scientific reliability.). Here, Dr. Hakalas methodology is so unreliable that he

cannot even reproduce his own results.

Professor Stulz has demonstrated that Dr. Hakala repeatedly and inexplicably changed

three key aspects of his AOL event study each time he has performed it. (Stulz Rpt. 102-109.)

First, as noted above (supra Part I.B.1), Dr. Hakalas identification of material events (i.e.,

AOL-specific news days), and his corresponding use of dummy variables for such days, has

19 Dr. Hakala has repeatedly insisted that an article by Aktas, de Bodt and Cousin in the Journal ofCorporate Finance (Aktas) supports his use of dummy variables. (Gesser Decl. Ex. 21 (Nihat Aktas et al., Event

Studies with a Contaminated Estimation Period, 13 J. Corp. Fin. 129 (2007)), cited in Hakala Rpt. 33 n.17; GesserDecl. Ex. 5 (Rebuttal Report of Scott. D. Hakala, dated July 17, 2008 (Hakala Rebuttal Rpt.)) 38 nn.19-20.)However, Defendants have already shown that Aktas neither follows nor recommends Dr. Hakalas approach.Unlike Dr. Hakala, Aktas (1) uses a statistical technique the two-state market model and not dummy variables tocontrol for contaminating events and (2) simulates a typical situation, in which a 225-day estimation period iscontaminated by three events (1.3%) nowhere near the 55% of days that Dr. Hakala dummies out in his currentreport. (See, e.g., Stulz Rpt. 99 n.67; Defs. Daubert Sur-Reply 4-5 (citing Aktas, supra, at 130, 137-38).)Moreover, Professor Stulz has shown that the point of the Aktas article is to control for event-induced varianceincreases, which Dr. Hakalas use of dummy variables fails to control for. (Stulz Rpt. 99 n.67.)

24



32/61

markedly increased over the course of his three AOL event studies, from 142 material events

(114 of which correspond to the Class Period here) in his first event study in In re AOL

Securities & ERISA Litigation (In re AOL), to 161 material events/dummy variables (132

within the Class Period) in his class certification declaration, to 211 days in his current report.

(See Stulz Rpt. 104 & n.73; Hakala Decl. Ex. B; Gesser Decl. Ex. 1 (Affidavit of Scott D.

Hakala filed in In re AOL, MDL No. 1500 (S.D.N.Y. filed Nov. 2, 2004) (Hakala Aff.)) App.

A 31.) Thus, the number of material events in the Class Period has jumped by 85% since Dr.

Hakalas Affidavit, and by nearly 60% since class certification in this case. (See Stulz Rpt.

104.)

Dr. Hakala has essentially admitted that no other economist could exactly duplicate his

selection of material events. As he has stated, it would not surprise him if two different

economists disagreed over the selection of 20-30 out of the 161 material days (12%-18%) he

identified at class certification. (Hakala 2007 Tr. 203:5-9.) At other times, Dr. Hakala has

suggested a lower error rate: you probably have a plus or minus error rate on event selection

about 5 to 10 percent. (Id. 203:23-25; Hakala 2008 Tr. 191:21-192:3 (suggesting, at most, a

5% rate).) But regardless of whether Dr. Hakala believes the error rate is 18%, 10% or 5%, his

estimates are dwarfed by the actual 85% change across his three AOL reports.20

(See Stulz Rpt.

20 Dr. Hakala claims that he made the changes because his opinions in In re AOL did not focus specificallyon the impact of analyst reports on AOLs share price, which is more specific to the issues in this case. (HakalaRebuttal Rpt. 35.) But this justification is inconsistent with what he says in his reports. In all three event studies,Dr. Hakalas purpose in identifying material events is the same: to control for all days when potentially material

information [about AOL] came into the market. (Hakala Aff. App. A 31; see also Hakala Rpt. 30 (verbatim);Hakala Decl. 12 (verbatim).) In all three event studies, Dr. Hakala specified the same protocol for identifying thepotentially material events, which included analyst reports: I relied upon the NASDAQ guidelines for materialnews as set forth and accepted by the SEC. The information considered included information in analysts reports,press releases, securities filings and published news articles (newspapers and daily publications, as well as moregeneral publications). (Hakala Aff. Ex. A 31 (emphasis added); see also Hakala Rpt. 30 & n. 14 (same); HakalaDecl. 12 & n.6 (same).) Moreover, Dr. Hakala has repeatedly stated that his identification of potentially materialevents and his use of dummy variables is supposed to control even for news unrelated to the subject of interest.(Hakala Rpt. 33 n.17 (emphasis added); id. (dummy variables used to control for AOL-specific news both relatedand unrelated to the allegations in this case); Hakala Decl. 15 n.9 (same).) Thus, [i]f [Dr. Hakalas] purpose of

25



33/61

104 n.74.)

Second, Dr. Hakala has inexplicably altered the estimation window he uses for his

regression. (Stulz Rpt. 108-109.) In In re AOL, Dr. Hakala used a period from January 12,

2001 through September 16, 2002, and specifically extended the study period beyond . . . July

26, 2002, which, he claimed, had increased the reliability and precision of the resulting

estimates. (Hakala Aff. Ex. A 30, 32.) At class certification, Dr. Hakala used a very similar

estimation window (January 9, 2001-September 16, 2002). (Hakala Decl. Ex. B.) However, in

his current report, Dr. Hakala uses the period from January 12, 2001-July 25, 2002 exactly the

one he rejected as too short in In re AOL. (See Stulz Rpt. 108-09.)

Third, Dr. Hakala has changed the market index and the components of the industry

indices he uses for his regression again, without providing any rationale for the change. 21

(Stulz Rpt. 105-107.)

Like Dr. Hakalas improper, post-hoc relevance determinations, his inadequately

explained and seemingly arbitrary changes to his event study have the effect of increasing the

statistical significance on days Dr. Hakala believes may be important to Plaintiffs claims. For

example, as a result of the changes that Dr. Hakala introduced since class certification, an

additional five days on which CSFB issued analyst reports (or the following days when the

reports were issued after the market closed), and an additional thirteen relevant curative

dummying out days is, as he claims, to obtain a set of days that are not contaminated by news, that set of dummyvariables should not change based on whether the defendants in the case are analysts or not. (Stulz Rpt. 104

n.74.)

21 In his rebuttal, Dr. Hakala claims that his latest selection of companies was not arbitrary, because theywere reported by analysts and in news reports as moving in concert with or otherwise related to AOL TimeWarners two business segments. (Hakala Rebuttal Rpt. 47.) But that is just the point: Professor Stulz showsthat some of the changes Dr. Hakala made were not supported by analyst reports or AOLs identification of its owncompetitors in its securities filings. (Stulz Rpt. 106.) Dr. Hakalas other justification, that his changes werenecessitated by the shorter estimation period (Hakala Rebuttal Rpt. 48), just begs the question of why he changedthe estimation period back to one that he considered inadequate at the time he did his first event study in In re AOL.

26



34/61

disclosure or inflationary event days have become, according to Dr. Hakala, statistically

significant at the 95% confidence level. (Stulz Rpt. 102; id. at Ex. 3.) Dr. Hakala undertook a

similar, results-oriented exercise in between his first AOL event study in In re AOL and his class

certification event study in this case by adding material event dates/dummy variables for days on

which Defendants issued their AOL reports, including, for example, September 19, 2001 (Hakala

2007 Tr. 86:17-24 (admitting that September 19, 2001 was not a materal event day in his In re

AOL event study), one of only two or three dates on which Dr. Hakala believes that a CSFB

report measurably increased AOLs stock price (id. 227:8-228:5; Hakala Rebuttal Rpt. 7, 32).

The arbitrary changes that Dr. Hakala made to his selection of material events and his

event study in order to increase statistical significance render his report unreliable. They do not

qualify as a methodology that could be used by another analyst to duplicate Dr. Hakalas

results. As a result, these changes standing alone would provide more than sufficient

grounds for excluding Dr. Hakalas opinions.

But these are not Dr. Hakalas only departures from standard event study methods that

make his results impossible to replicate. Indeed, Dr. Hakalas unique and arbitrary take on

statistical significance flies in the face of basic scientific methodology. Dr. Hakala at times

eschews statistical significance entirely. Thus, as Professor Stulz has pointed out:

[i]t is contrary to established scientific methodology to assume thata day associated with insignificant abnormal returns can beconsidered to be curative . . . . However, almost half of the daysDr. Hakala claims are relevant corrective disclosures ofadvertising-related information are insignificant even in his model,which systematically overstates the significance of abnormalreturns.

(Stulz Rpt. 58; id. 58 n.47 (noting that 23 of 49 days cited as relevant in Dr. Hakalas Ex.

C-1a are statistically insignificant).) The insignificant dates on which Dr. Hakala attributes

inflation or loss causation to Defendants include January 12, 2001, the first day of the Class

27



35/61

Period, when Defendants allegedly issued their first AOL report, which contained an allegedly

false $80 price target, and July 18-19, 2002, the days the Washington Post published articles

raising questions about AOLs accounting for certain advertising transactions and, according to

Dr. Hakala, the first corrective disclosures of the alleged fraud relating to advertising

investigations. (See Hakala Rpt. Ex. B-1, at 1, 11.) Nothing in Dr. Hakalas report explains why

he believes it is appropriate to jettison statistical significance at some times but not others, nor is

there any basis from which another economist could duplicate his unscientific significance

determinations. As a result, Dr. Hakalas opinions lack any indicia of reliability, and must be

rejected.

C. Dr. Hakalas Use of a Proxy for the Effect of DefendantsStatements Is Wholly Unreliable

Besides Dr. Hakalas repetition of the flaws that led Judge Zobel to exclude his opinions

in Xcelera, his latest event study contains an equally fatal problem. Dr. Hakala does not measure

the price inflation attributable to CSFBs alleged misstatements by analyzing the movement in

AOLs share price when Defendants actually issued their AOL reports as he should have under

established law and economic principles. Rather, he attempts to do so based on AOLs average

price movement on 23 days when CSFB did not speak but when other analysts issued negative

reports regarding information that CSFB is not alleged to have known. (Hakala Rpt. 14, 27-

28.) Dr. Hakala then takes the average AOL price movement on such days which he computes

to be approximately -2.7% and introduces that amount (or a percentage of that amount) of

artificial inflation into AOLs stock price on certain days on which CSFB did issue an AOL

report, regardless of AOLs actual stock price movement on those days. Dr. Hakala thus relies

on the market impact of other analysts statements on AOLs share price as a proxy for the

purported impact of CSFBs statements, in an effort to establish that, had Defendants made

28



36/61

disclosures that were different from what Defendants actually published in their reports, these

statements would have affected AOLs stock price. (Id.; see also Stulz Rpt. 35.) Dr. Hakalas

approach is inherently and profoundly speculative and it is inconsistent with generally accepted

methodologies in the field of financial economics. (See Defs. S.J. Mem. 71-72; Stulz Rpt.

5(d), 32-48.)

There are at least three serious methodological flaws in Dr. Hakalas use of an analyst

proxy that render his opinion inadmissible. First, Dr. Hakalas proxy has absolutely nothing to

do with AOLs actual stock price movements. Although Dr. Hakala performed what he

considers to be a detailed event study, he often disregards the results of that study and instead

replaces them with the unrelated computations he derived from his proxy analysis. Thus, on

January 12, 2001, the day CSFB issued its first AOL report during the Class Period, Dr. Hakalas

event study calculated a negative abnormal return for AOL of -2.23%. (Hakala Report Ex. B-1,

at 1.) Dr. Hakala, however, applies his proxy to introduce positive 2.7% artificial inflation into

AOLs stock price that day. (Id. Ex. C-1, at 1.) In other words, Dr. Hakala seems to think that

on a day when AOLs stock price declined, its price would have been exactly 2.7% lower but for

Defendants report. Yet Dr. Hakala has never identified any academic support for this approach

of substituting his hypothetical proxy for his actual event study results, and none exists.

Indeed, Dr. Hakalas use of his analyst proxy, rather than his event studys results, to

demonstrate inflation and market impact is virtually identical to the approach that the Southern

District of New York excluded as so transparently unreliable as to be inadmissable as a matter

of law in DeMarco v. Lehman Brothers, Inc., 222 F.R.D. 243, 248 (S.D.N.Y. 2004). There, the

plaintiffs expert relied on an analyst proxy that was based on an academic study, which had

found that significant ratings downgrades by research analysts result, on average, in abnormal

29



37/61

returns of -5%. Id. at 248-49. Yet the court rejected the study as a basis for the experts opinion,

finding, among other things, that the proxy has virtually nothing to do with [the defendant

analyst], [the analysts] alleged influence on the market (which [the expert] largely just

assumes), or [the analysts] actual statements, and appears irrelevant on its face. Id. at 248. In

fact, Dr. Hakalas approach is even less reliable than the approach that the DeMarco court

rejected, because Dr. Hakalas average analyst effect is computed using far fewer data points

over a much shorter period of time. See id. (noting that study looked at ratings changes for 200

companies over three-year period).

Second, Dr. Hakalas approach is intellectually unsound because it essentially assumes

that any time one analyst issues a negative report about a company, it will have an effect that is

very similar to a different analysts negative report concerning different information about that

company. Dr. Hakalas approach thus contradicts the literature in financial economics that

shows that the impact of analyst statements on stock price, when there is one, depends on a wide

variety of factors for which Dr. Hakala does not control.22

(Stulz Rpt. 5(d)(i); see id. 37;

cf. Stulz Decl. 32-33, 36.)

Notably, Dr. Hakala never bothers to compare the information disclosed on his 23 proxy

days and the negative information that CSFB allegedly should have disclosed but did not. (Stulz

Rpt. 35-36; Hakala 2008 Tr. 95:14-96:15 (acknowledging that his opinion does not make any

effort to account for the varying content and importance of the news released on various days, or

different ratings, estimates and price targets that other analysts had on AOL).) As a result, 11 of

Dr. Hakalas 23 analyst proxy days contain commentary and estimates for AOLs 2003

22 Such factors include, but are not limited to, the content of the report, the ranking of the analyst,statements by other analysts and the subject company here, AOL and the timing of the recommendation orearnings forecast. (See Stulz Rpt. 36.)

30



38/61


39/61

were dates when confounding news from the issuer was minimal. 222 F.R.D. at 249. Such

an approach which Dr. Hakala also followed here simply is not admissible under Daubert or

Fe

in re credit suisse_304_defendant's brief re motion to preclude expert opinions

Documents